x********************************* * * * Changes in islasso * * * ********************************* =============== version 1.5.1 =============== * Some bugs fixed. =============== version 1.5.0 =============== * Some bugs fixed. Other S3 methods implemented. =============== version 1.4.3 =============== * Some bugs fixed. =============== version 1.4.2 =============== * Some bugs for binomial family fixed. =============== version 1.4.1 =============== * Some bugs fixed. =============== version 1.4.0 =============== * New optimization algorithm for the 'islasso' method. The algorithm is now stable for all the implemented distributions. * In aic.islasso function the available methods are "AIC", "BIC", "AICc", "eBIC", "GCV", "GIC". * New class of functions named 'islasso.path' have been created. The main function 'islasso.path()' allows to build the coefficient profile for a fixed sequence of lambda values. * A new function GoF.islasso.path() allows to extract the optimal value of the tuning parameter which minimizes a fixed criterion. Available criteria are the same as in aic.islasso function. * Some bugs fixed. =============== version 1.3.1 =============== * Some bugs fixed. =============== version 1.3.0 =============== * Vignette added to the package 'islasso' * Some bugs fixed. =============== version 1.2.3 =============== * Some bugs fixed. =============== version 1.2.2 =============== * Some bugs fixed. =============== version 1.2.1 =============== * Some bugs fixed. =============== version 1.2.0 =============== * New implementation of the estimating algorithm. Now islasso is much stabler and faster. * New function: general linear hypotheses for linear combinations of the regression coefficients, including confidence intervals. * New changes: - prediction function includes confidence intervals for the fitted values - step Halving with Armijo's rule has been improved. - convergence criterion has been improved * Some bugs fixed. =============== version 1.1.0 =============== * New implementation of the estimating algorithm. Now islasso is much stabler and faster reducing the number of iterations to reach convergence. * New changes: - step Halving with Armijo's rule has been implemented. - the algorithm includes the possibility to use the elastic-net approach defining an alpha parameter in the objective function as in glmnet package. - the summary method includes now the degree of freedom for each covariate, and in addition it is possible to choose between t-test or z-test (only for gaussian family). - optim.islasso has been renamed as aic.islasso the interval specification is not required now, the new select the best tuning parameter based on the minimization of the AIC or the BIC. - the function islasso.control has been renamed as is.control and some control parameters have been modified. - two trace version has been implemented in the function is.control a compact version (trace=1) and a long version (trace=2). * Some bugs fixed.